I am not surprised …
SAS XPT does NOT support any characters other than US-ASCII, so also no Japanese characters.
So, people use tricks (workarounds) to enforce Japanese characters, taking two bytes instead of one for a character.
However, any receiving software, including P21 cannot know nor detect what the encoding of the characters in the XPT file is, also as this information is not present in the header of the XPT files.
Bytes are just bytes, and no software can know whether a byte represents an ASCII or (part of) a Japanese character (SHIFT-JIS?)
So, what you ask for is essentially not possible.
Define-XML however, as a modern standard, uses Unicode and UTF-8 encoding.
The only thing you can do is to load your define.xml in a software (can be a simple XML editor that supports UTF-8) and manually write the Japanese into the right XML elements.
There is also very user-friendly, GUI-based software on the market that allows you to do so without XML knowledge.
It is high time that regulatory authorities get rid of XPT and move to CDISC Dataset-JSON format.
You can always mail me for more explanation and advice. My mail address is easy to find out.